Curvature-difference localization using underfitted baselines identifies overfitting-driven memorization in diffusion models and outperforms attention-based methods on Stable Diffusion.
However, crucially, the curvatureκ1 at this mode does not stop increasing; it continues to rise significantly as training proceeds to the final step (60,000 steps)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Localizing Memorized Regions in Diffusion Models via Coordinate-Wise Curvature Differences
Curvature-difference localization using underfitted baselines identifies overfitting-driven memorization in diffusion models and outperforms attention-based methods on Stable Diffusion.